Improving Text-to-Music Generation with Human Preference Rewards

research work·research paper·active

Research work examining music generation, text-to-music generation.

Recorded facts

Official site	https://arxiv.org/abs/2606.21670 ↗
Geography	Global
arxiv	2606.21670
venue	ICME 2026 Grand Challenge on Academic Text-to-Music Generation
authors	Yonghyun Kim; Junwon Lee; Haiwen Xia; Yinghao Ma; Chris Donahue
methods	unknown
code urls	unknown
demo urls	unknown
exact title	Improving Text-to-Music Generation with Human Preference Rewards
project urls	unknown
original title	unknown
citation counts	unknown
research topics	music generation; text-to-music generation
peer review status	not established from abstract metadata
disclosed conflicts	unknown
stated contribution	Presents or evaluates the system, method, benchmark, or analysis identified in the paper title.
us market scope basis	Included as a materially relevant public research artifact in the US-facing AI-music ecosystem; direct affiliation varies.
funding acknowledgements	unknown
affiliations at publication	Georgia Institute of Technology; Carnegie Mellon University
publication or preprint date	2026-06-19
abstract level neutral summary	The work studies music generation, text-to-music generation; methods and evaluation details are in the official abstract.
datasets benchmarks models tools used	CLAP
correction withdrawal retraction status	No withdrawal marker observed in captured arXiv metadata.